CN107154044A - A kind of dividing method of Chinese meal food image - Google Patents

Abstract

The method that the present invention is provided realizes the segmentation to image by gathering the texture image progress subsequent treatment of Chinese meal food image, a variety of characteristics of image need not be gathered during segmentation, and the accuracy rate of Chinese meal food image segmentation can be improved using this method, so as to help the identification of Chinese meal food image.

Description

A Segmentation Method of Chinese Food Image

technical field

The present invention relates to the technical field of image processing, and more particularly, to a method for segmenting Chinese food images.

Background technique

The classic image segmentation method is based on the characteristics of the image. The main function of the image segmentation method is to segment the region with the same or similar characteristics. According to the different characteristics used, it can be roughly divided into the following categories:

a). Based on the segmentation method of threshold value processing, it can divide the pixels of the image into different categories by setting the basic feature threshold. Commonly used features include: grayscale features, color features, or features transformed from original grayscale values or color values, etc. An obvious way to extract objects or foreground objects from the background is to choose an appropriate threshold T to separate these feature patterns.

b). A segmentation method based on edge detection. Edge detection is the most commonly used method to segment images according to grayscale mutations. The edge is a collection of pixels on the boundary line of the image, which indicates the discontinuity of the local features of the image, and reflects the sudden change of image features such as grayscale features, color features, and texture features of the image. For example, on both sides of a step-shaped edge, the gray value of the pixel point will have a significant difference, while at the roof-shaped edge, the gray value shows a steep rise or fall.

c). Segmentation method based on regional features: This method is based on the similarity criterion of pixels in the same region of the image for segmentation. This method performs similarity clustering on the values in the feature space of each pixel, and takes into account the spatial domain information of each pixel, and then segments the target area in the image. Commonly used methods include seed region growth, region splitting and aggregation, and morphological watershed methods. However, because the similarity threshold is not easy to control, the result obtained by the segmentation method using regional features is not smooth enough in the boundary area.

d). Segmentation methods based on edge features and regional features: There are deficiencies in the segmentation methods that use edge features or regional features alone, so some researchers put forward the Some improved models, such as segmentation methods based on variational models and segmentation methods based on graph theory, etc.

Due to the different image categories, the image segmentation methods used are also different. In the segmentation method of food images, when using common color features and brightness features, it can well segment out the areas with more obvious and bright colors, but cannot segment out the color In the darker area, due to the diversity and complexity of ingredients in Chinese food images, it is necessary to use a variety of image features as feature comparisons for image segmentation, so as to segment a complete food area and remove the background area.

Contents of the invention

In order to solve the above-mentioned defects in the prior art that multiple different image features need to be used for comparison when segmenting the dark region of the Chinese food image, the present invention provides a segmentation method of the Chinese food image. The method collects the Chinese food image Subsequent processing is performed on the texture image to achieve image segmentation. During the segmentation process, multiple image features do not need to be collected, and the application of this method can improve the accuracy of Chinese food image segmentation, thereby helping the recognition of Chinese food images.

For realizing above-mentioned purpose of the invention, the technical scheme that adopts is:

A method for segmenting Chinese food images, comprising the following steps:

S1. Use the texture enhancement filter to filter the captured Chinese food image under m different scale parameters to obtain the texture image of the image under m different scale parameters; the value range of m is 8-16;

S2. Calculate the mean values of the 16 texture images obtained in step S1, and use the calculated mean values as thresholds to binarize the corresponding texture images to obtain the foreground and background regions of the texture images under threshold conditions;

S3. For each texture image, the center point of its foreground area is calculated separately, so as to be used as the position for placing the Gaussian function, and k times the number of pixels contained in the foreground area is used as the standard deviation to construct a corresponding Gaussian mask function, wherein The value of k ranges from 0.3 to 0.5; multiply the obtained 16 Gaussian mask functions by the corresponding weight parameters and add them up to obtain the final Gaussian mask;

S4. Multiply the obtained Gaussian mask with the texture image produced by the Chinese food image when the scale parameter of the texture enhancement filter is [0.5m], and record the result as graph G, where [0.5m] means 0.5m for rounding operation; use the SLIC method to perform superpixel segmentation on the graph G, after segmentation, get the category of the block to which each pixel in the image belongs, which is called the label matrix L, and record L as the label map of the graph G ;

S5. Calculate the average value Gk of the pixel area with the same category mark for each pixel in the graph G, compare the average value Gk with the overall average value Gu of the graph G, if Gk>Gu, then the pixel area with the same mark The pixel value of each pixel is set to 1, and the pixel area with the same label is marked as the foreground area, otherwise, the pixel value of each pixel point in the pixel area with the same label is set to 0, and the pixel area with the same label Mark as background area;

S6. Perform morphological opening and closing operations on the foreground area and the background area to smooth the edge areas of the foreground area and the background area, and then segment the foreground area and the background area.

Preferably, the texture enhancement filter is a Gabor function.

Compared with prior art, the beneficial effect of the present invention is:

The method provided by the present invention realizes image segmentation by collecting texture images of Chinese food images for subsequent processing, and does not need to collect multiple image features during the segmentation process, and the application of this method can improve the accuracy of Chinese food image segmentation, thereby helping in the recognition of Chinese food images.

Description of drawings

Figure 1 is a schematic flow chart of the method.

Fig. 2 is the texture image produced by the Chinese food image when the scale parameter of the texture enhancement filter is 8.

Figure 3 is a schematic diagram of segmentation.

detailed description

The accompanying drawings are for illustrative purposes only and cannot be construed as limiting the patent;

The present invention will be further elaborated below in conjunction with the accompanying drawings and embodiments.

Example 1

As shown in Figure 1, the method provided by the present invention specifically includes the following steps:

S1. Use the texture enhancement filter to filter the captured Chinese food image under m different scale parameters to obtain the texture image of the image under m different scale parameters; the value range of m is 8-16;

S2. Calculate the mean values of the 16 texture images obtained in step S1, and use the calculated mean values as thresholds to binarize the corresponding texture images to obtain the foreground and background regions of the texture images under threshold conditions;

S3. For each texture image, the center point of its foreground area is calculated separately, so as to be used as the position for placing the Gaussian function, and k times the number of pixels contained in the foreground area is used as the standard deviation to construct a corresponding Gaussian mask function, wherein The value of k ranges from 0.3 to 0.5; multiply the obtained 16 Gaussian mask functions by the corresponding weight parameters and add them up to obtain the final Gaussian mask;

S4. Multiply the obtained Gaussian mask with the texture image produced by the Chinese food image when the scale parameter of the texture enhancement filter is [0.5m], and record the result as graph G, where [0.5m] means 0.5m for rounding operation; use the SLIC method to perform superpixel segmentation on the graph G, after segmentation, get the category of the block to which each pixel in the image belongs, which is called the label matrix L, and record L as the label map of the graph G ;

S5. Calculate the average value Gk of the pixel area with the same category mark for each pixel in the graph G, compare the average value Gk with the overall average value Gu of the graph G, if Gk>Gu, then the pixel area with the same mark The pixel value of each pixel is set to 1, and the pixel area with the same label is marked as the foreground area, otherwise, the pixel value of each pixel point in the pixel area with the same label is set to 0, and the pixel area with the same label Mark as background area;

S6. Perform morphological opening and closing operations on the foreground area and the background area to smooth the edge areas of the foreground area and the background area, and then segment the foreground area and the background area. A schematic diagram of the foreground region obtained after segmentation is shown in Figure 3.

In this embodiment, the texture enhancement filter is a Gabor function. The Gabor function is a linear filter used for edge extraction. Its frequency and direction expression are similar to the human visual system. Therefore, the texture of the original image at different scales and in different directions can be extracted by using the Gabor filter. The mathematical expression of the two-dimensional Gabor function is

Where x'=x cosθ+y sinθ, y'=-x sinθ+y cosθ

In this embodiment, x and y are two-dimensional random variables. According to the composition of the smallest particles in the Chinese food image, the window size of the Gabor filter is set to 32*32, and the parameter λ ranges from 1 to 16, a total of 16 scales , the parameter θ is set to four directions of 0°, 45°, 90°, and 135°, and the phase is 0, the standard deviation σ is 2π, the aspect ratio γ is 0.5, and the texture image features extracted when the parameter λ of the filter is 8 are shown in Figure 2.

Apparently, the above-mentioned embodiments of the present invention are only examples for clearly illustrating the present invention, rather than limiting the implementation of the present invention. For those of ordinary skill in the art, on the basis of the above description, other changes or changes in different forms can also be made. It is not necessary and impossible to exhaustively list all the implementation manners here. All modifications, equivalent replacements and improvements made within the spirit and principles of the present invention shall be included within the protection scope of the claims of the present invention.

Claims (2)

1. a segmentation method of Chinese food image, is characterized in that: comprise the following steps: S1. Use the texture enhancement filter to filter the captured Chinese food image under m different scale parameters to obtain the texture image of the image under m different scale parameters; the value range of m is 8-16; S2. Calculate the mean values of the 16 texture images obtained in step S1, and use the calculated mean values as thresholds to binarize the corresponding texture images to obtain the foreground and background regions of the texture images under threshold conditions; S3. For each texture image, the center point of its foreground area is calculated separately, so as to be used as the position for placing the Gaussian function, and k times the number of pixels contained in the foreground area is used as the standard deviation to construct a corresponding Gaussian mask function, wherein The value of k ranges from 0.3 to 0.5; multiply the obtained 16 Gaussian mask functions by the corresponding weight parameters and add them up to obtain the final Gaussian mask; S4. Multiply the obtained Gaussian mask with the texture image produced by the Chinese food image when the scale parameter of the texture enhancement filter is [0.5m], and record the result as graph G, where [0.5m] means 0.5m for rounding operation; use the SLIC method to perform superpixel segmentation on the graph G, after segmentation, get the category of the block to which each pixel in the image belongs, which is called the label matrix L, and record L as the label map of the graph G ; S5. Calculate the average value Gk of the pixel area with the same category mark for each pixel in the graph G, compare the average value Gk with the overall average value Gu of the graph G, if Gk>Gu, then the pixel area with the same mark The pixel value of each pixel is set to 1, and the pixel area with the same label is marked as the foreground area, otherwise, the pixel value of each pixel point in the pixel area with the same label is set to 0, and the pixel area with the same label Mark as background area; S6. Perform morphological opening and closing operations on the foreground area and the background area to smooth the edge areas of the foreground area and the background area, and then segment the foreground area and the background area. 2. the segmentation method of Chinese food image according to claim 1, is characterized in that: described texture enhancement filter is Gabor function.